For many years it has been a goal of scientists in the field of specifically targeted drug therapy that antibodies could be used for the specific delivery of chemotherapy drugs to human cancers. Realization of such a goal could finally bring to cancer chemotherapy the concept of the magic bullet. A significant advance toward achieving this goal came with the advent of the hybridoma technique of Kohler and Milstein in 1975, and the subsequent ability to generate monoclonal antibodies (MAbs). During the past 25 years monoclonal antibodies have been raised against many antigenic targets that are over-expressed on cancerous cells. Either alone, or as conjugates of drugs, toxins, radionuclides or other therapy agents, many antibodies have been tested pre-clinically, and later in clinical trials.
Generally, antibodies by themselves, often termed “naked antibodies,” have not been successful at making long-term survivorship the norm in patients with solid tumors, although survival advantages have lately been seen with antibody treatments directed against both breast and colon cancer (antibodies against HER2-neu and 17-1A, respectively). With hematological malignancies more success is being achieved with naked antibodies, notably against the B-cell lymphomas (antibodies against CD20 and CD22 on the surface of B-cells).
It appears self-evident, however, that the use of conjugates of tumor-associated antibodies and suitable toxic agents will be more efficacious than naked antibodies against most clinical cases of cancer. Here, an antibody also carries a toxic agent specifically to the diseased tissue, in addition to any toxicity it might inherently have by virtue of natural or re-engineered effector functions provided by the Fc portion of the antibody, such as complement fixation and ADCC (antibody dependent cell cytotoxicity), which set mechanisms into action that may result in cell lysis. However, it is possible that the Fc portion is not required for therapeutic function, as in the case of antibody fragments, other mechanisms, such as apoptosis, inhibiting angiogenesis, inhibiting metastatic activity, and/or affecting tumor cell adhesion, may come into play. The toxic agent is most commonly a chemotherapy drug, a particle-emitting radionuclide, or a bacterial or plant toxin. Each type of conjugate has its own particular advantages. Penetrating radionuclides and the bacterial and plant toxins are extremely toxic, usually orders of magnitude more toxic than standard chemotherapy drugs. This makes the former two useful with antibodies, since in a clinical situation the uptake of antibodies into diseased tissue is extremely low. The low antibody tumor uptake in clinical practice and the relatively low toxicity profile of cancer chemotherapy drugs, combined, is a major reason why antibody-drug conjugates have failed to live up to their promise, to date.
In preclinical animal xenograft models, set up to study human cancer, many antibody conjugates have been described which are able to completely regress or even cure animals of their tumors. However, tumor uptakes of antibody conjugates in many of these animal xenograft models are often in the 10-50% injected dose per gram of tissue range, whereas in the clinical situation, tumor uptakes in the 0.1-0.0001% injected dose per gram of tissue are more normal. It is no surprise, then, that antibody conjugates made with the more toxic radionuclides and toxins have generally fared somewhat better, clinically, than the corresponding antibody-drug conjugates with standard chemotherapeutic drugs. However, radionuclide antibody conjugates can often produce significant toxicity due to the presence of excess circulating, decaying radioactivity compared to tumor-localized activity. Toxin-antibody conjugates have suffered from the dual drawbacks of non-target tissue damage and immunoreactivity toward the plant or bacterial protein that is generally used. Whereas antibodies can now be made in human or in humanized (complementarity-determining region-grafted) forms, de-immunization of the toxin part of any conjugate will likely remain a significant obstacle to progress.
Despite the lack of efficacy in a clinical setting seen to date, antibody-drug conjugates still have compelling theoretical advantages. The drug itself is structurally well defined, not present in isoforms, and can be linked to the antibody protein using very well defined conjugation chemistries, often at specific sites remote from the antibodies' antigen binding regions. MAb-drug conjugates can be made more reproducibly than chemical conjugates involving antibodies and toxins, and, as such, are more amenable to commercial development and regulatory approval. For such reasons, interest in drug conjugates of antibodies has continued despite the disappointments encountered. In some recent instances, however, preclinical results have been quite promising. With continuing refinements in conjugation chemistries, and the ability to remove or reduce immunogenic properties of the antibody, the elusive promise of useful antibody-drug conjugates for clinical cancer therapy are being newly considered.
Relevant early work on antibody-drug conjugates found during in vitro and in vivo preclinical testing that the chemical linkages used often resulted in the loss of a drug's potency. Thus, it was realized many years ago that a drug would ideally need to be released in its original form, once internalized by a target cell by the antibody component, in order to be a useful therapeutic. Work during the 1980s and early 1990s then focused largely on the nature of the chemical linker between the drug and the antibody. Notably, conjugates prepared using mild acid-cleavable linkers were developed, based on the observation that pH inside tumors was often lower than normal physiological pH (U.S. Pat. Nos. 4,542,225; 4,569,789; 4,618,492; and 4,952,394). This approach culminated in a landmark paper by Trail et al. (Science 261:212-215 (1993)) showing that antibody-doxorubicin (DOX) conjugates, prepared with appropriate linkers, could be used to cure mice bearing a variety of human tumor xenografts, in preclinical studies.
This result was achieved with an antibody (termed BR96) that had a very large number of receptors on the tumor cells being targeted and the antibody-drug conjugate was highly substituted (6-8 DOX residues per unit of antibody). However, the conjugate was given in massive doses on a repeat basis in order to achieve efficacy. A need exists for drug-antibody conjugates that are efficacious at much lower dosages.
In the clinical situation, tumor uptake of antibodies would be much lower, and it has been suggested that more toxic drugs are needed to achieve a desirable therapeutic effect. More toxic drugs were used in the development of several distinct antibody-drug conjugates (U.S. Pat. Nos. 5,208,020; 5,416,064; 5,877,296; and 6,015,562). These efforts use drugs, such as derivatives of maytansinoids and calicheamicin, which are essentially too toxic to be used in standard chemotherapy. Conjugation to an antibody enables relatively more of the drug to be targeted to a tumor in relation to the often non-specific cell and protein binding seen with chemotherapy alone. The exquisite toxicity of drugs such as these might overcome the low levels of tumor-targeted antibody seen clinically, due to the low level of antigen binding sites generally seen on tumor targets. In preclinical studies, cures of mice bearing human tumor xenografts were seen at much lower doses of antibody-drug conjugate, than seen previously with antibody-drug conjugates using standard drugs, such as DOX (Liu et al., Proc. Natl. Acad. Sci. USA 93:8616-8623 (1996) and Hinman et al., Cancer Res. 53:3336-3342 (1993)). In the case of the maytansinoid-antibody conjugates (Liu), the amount of conjugate needed for therapy was over >50-fold less than needed previously with DOX conjugates (Trail, supra).
During development of these conjugates the linker between drug and antibody was thought to be critical for retention of good anti-tumor activity both in vitro and in vivo. The cited conjugates were made with an intracellularly-cleavable moiety (hydrazone) and a reductively labile (disulfide) bond between the drug and the antibody. While the hydrazone bond is apparently stable to in vivo serum conditions, normal disulfide bonds were found to be not stable enough for practical use. Conjugates were made that replaced a standard disulfide linkage with a hindered (geminal dimethyl) disulfide linkage in the case of the calicheamicins, or a methyl disulfide in the case of the maytansinoids. While this work was being done, separate work also continued on newer anthracycline-substituted antibody conjugates. In the case of newer DOX conjugated antibodies, it was found that superior results could be obtained by incorporating just a hydrazone as a cleavable unit, and attaching DOX to antibody via a thioether group, instead of a disulfide (U.S. Pat. No. 5,708,146). When linked in such a manner, and also using a branched linker capable of doubling the number of DOX units per MAb substitution site, an approximate order of magnitude increase in the efficacy of the new DOX-MAb conjugates was obtained (King et al., Bioconjugate Chem. 10:279-288, (1999)).